sigma08-007. 269KB Jun 04 2011 12:10:12 AM
Symmetry, Integrability and Geometry: Methods and Applications
SIGMA 4 (2008), 007, 14 pages
Three Order Parameters in Quantum XZ
Spin-Oscillator Models with Gibbsian Ground States⋆
Teunis C. DORLAS
†
and Wolodymyr I. SKRYPNIK
‡
†
Dublin Institute for Advanced Studies, School of Theoretical Physics,
10 Burlington Road, Dublin 4, Ireland
E-mail: dorlas@stp.dias.ie
‡
Institute of Mathematics of NAS of Ukraine, Kyiv, Ukraine
E-mail: skrypnik@imath.kiev.ua
Received October 29, 2007, in final form January 08, 2008; Published online January 17, 2008
Original article is available at http://www.emis.de/journals/SIGMA/2008/007/
Abstract. Quantum models on the hyper-cubic d-dimensional lattice of spin- 21 particles
interacting with linear oscillators are shown to have three ferromagnetic ground state order
parameters. Two order parameters coincide with the magnetization in the first and third
directions and the third one is a magnetization in a continuous oscillator variable. The
proofs use a generalized Peierls argument and two Griffiths inequalities. The class of spinoscillator Hamiltonians considered manifest maximal ordering in their ground states. The
models have relevance for hydrogen-bond ferroelectrics. The simplest of these is proven to
have a unique Gibbsian ground state.
Key words: order parameters; spin-boson model; Gibbsian ground state
2000 Mathematics Subject Classification: 82B10; 82B20; 82B26
1
Introduction
In this paper we consider quantum lattice models of oscillators interacting with spins whose variables are indexed by the sites of a hyper-cube Λ with the finite number of sites |Λ| in the
hyper-cubic lattice Zd . Interaction is considered to be short-range and translation invariant.
The corresponding Hamiltonian HΛ is expressed in terms of the oscillators variables qΛ =
(qx , x ∈ Λ) ∈ R|Λ| and spin 12 Pauli matrices SΛl = (Sxl , x ∈ Λ, l = 1, 3), defined in the tensor product of the 2|Λ| -dimensional Euclidean space and the space of square integrable functions
L2Λ = (⊗E2 )|Λ| ⊗ L2 (R|Λ| ),
HΛ =
X
x∈Λ
X
1
JA S[A]
− ∂x2 + µ2 (qx + ηφx (SΛ3 ))2 − µ +
+ VΛ ,
µ ≥ 0,
η ∈ R,
(1.1)
A⊆Λ
where ∂x is the partial derivative in qx , JA and VΛ are real-valued measurable functions, the
first of which depends on qΛ and the second on SΛ3 , qΛ and the translation invariant φx is given
by (see Remark 2 in the end of the paper)
φx (sΛ ) =
X
J0 (x; A)s[A] ,
J0 (x; x) = 1.
(1.2)
A⊆Λ
⋆
This paper is a contribution to the Proceedings of the Seventh International Conference “Symmetry in
Nonlinear Mathematical Physics” (June 24–30, 2007, Kyiv, Ukraine). The full collection is available at
http://www.emis.de/journals/SIGMA/symmetry2007.html
2
T.C. Dorlas and W.I. Skrypnik
For products of operators, functions and variables we use the following notation: B[A] =
Q
Bx .
x∈A
The scalar product in (⊗E2 )|Λ| , L2Λ will be denoted by (·, ·)0 , (·, ·), respectively. The Schwartz
space of test functions on Rn will be denoted by S(Rn ).
We require that the Hamiltonian is well defined and bounded from below on (⊗E2 )|Λ| ⊗
C0∞ (R|Λ| ), i.e. the tensor product of the 2|Λ| -dimensional Euclidean space and the space of
infinitely differentiable functions with compact supports. The ground state average for an
observable (operator) F is determined by
hF iΛ = ZΛ−1 (ΨΛ , F ΨΛ ),
ZΛ = (ΨΛ , ΨΛ ) = ||ΨΛ ||2 ,
where ZΛ is a partition function. For partial cases of F we have
hˆ
q[A] φ[A′ ] (SΛ3 )iΛ
=
ZΛ−1
Z
q[A] (ΨΛ (qΛ ), φ[A′ ] (SΛ3 )ΨΛ (qΛ ))0 dqΛ ,
where the integration is performed over R|Λ| and qˆ[A] is the operator of multiplication by qx .
0 (s ) of the Euclidean space (⊗E2 )|Λ| , diagonaliWe will employ the orthonormal basis ψΛ
Λ
3
0 (s ) = ⊗
0
0
zing SΛ , which is chosen in the following way: ψΛ
Λ
x∈Λ ψ (sx ), sx = ±1, ψ (1) = (1, 0),
0
1
0
0
3
0
0
2
ψ (−1) = (0, 1), S ψ (s) = ψ (−s), S ψ (s) = sψ (s). For F ∈ LΛ we have the following
decomposition
F (qΛ ) =
X
0
(sΛ ),
F (qΛ ; sΛ )ψΛ
sΛ
where the summation is performed over the |Λ|-fold Cartesian product (−1, 1)|Λ| of the set (−1, 1).
The Hamiltonians in (1.1) are employed in hydrogen-bond ferroelectric crystal models, considered in [1, 2, 3], and describe interaction between heavy ions (oscillators with constant frequency) and protons (spins). The second term with JA = 0, |A| ≥ 2 corresponds to the energy
of protons, tunneling along hydrogen bonds from one well to another, and Jx is associated
with
P
3
J1 (A)S[A]
the tunneling frequency. The last term in the expression for our Hamiltonian VA =
A⊆Λ
describes many-body interaction between protons (J1 (A) is the intensity of the |A|-body interaction).
A rigorous analysis of a mean-field version of the
Hamiltonian in (1.1) with φx (SΛ3 ) given
P
in (1.2) that is linear in S 3 , Jx 6= 0, VΛ = −(µη)2
φ2x (SΛ3 ), JA = 0 for |A| ≥ 2 and J0 (x; y)
x∈Λ
uniformly in lattice sites proportional to |Λ|−1 was carried out in [3] in the framework of the Bogolyubov approximating Hamiltonian method [4] and occurrence of spin and oscillator orderings
(the corresponding order parameters are non-zero) for two-body interaction between protons at
non-zero temperatures was proved. To establish such orderings for ground states without the
mean-field limit in a general case is an important task for a theory.
The oscillator and spin orderings are established if one proves the existence of ferromagnetic
oscillator and spin long-range orders (lro’s) in ground states for the corresponding Hamiltonians.
This means that the ground state averages hˆ
qx qˆy iΛ , hSxj Syj iΛ , j = 1, 3 are bounded uniformly in Λ
from below by positive numbers . Occurrence of the ferromagnetic lro’s implies the existence
P of
l
−1
Sxl ,
the spin order parameters (magnetizations in the first and third directions) MΛ = |Λ|
x∈Λ
P
qx in the thermodynamic limit
l = 1, 3, and the oscillator order parameter QΛ = |Λ|−1
x∈Λ
(Λ → Zd ) since the ground state averages of their squares are uniformly bounded in Λ from
below by a positive number.
Three Order Parameters in Quantum XZ Spin-Oscillator Models
3
In this paper we find functions VΛ , depending on oscillator variables, for which ground states
or eigenstates ΨΛ of the Hamiltonians in (1.1) are Gibbsian
X 1
0
e− 2 U (sΛ ;qΛ ) ψΛ
ΨΛ (qΛ ) =
(sΛ )ψ0Λ (qΛ ),
(1.3)
sΛ
with the linear in qΛ spin-oscillator quasi-potential energy U
X
qx φx (sΛ ) + U 0 (sΛ ),
U (sΛ ; qΛ ) = 2µη
(1.4)
x∈Λ
where ψ0Λ (qΛ ) =
Q
x∈Λ
1
ψ0 (qx ), ψ0 (q) = (µπ −1 ) 4 exp{− µ2 q 2 }, is the ground state of the free
oscillator Hamiltonian, that is the first term in the right-hand side of (1.1) with η = 0. We
prove the maximal ordering in the corresponding systems (provided some simple conditions
on U 0 and φx are satisfied): magnetizations MΛl , l = 1, 3, QΛ are non-zero. No other ground
states are known with such the property.
Gibbsian ground states were introduced by Kirkwood and Thomas in [5] in XZ spin- 12 models
with Hamiltonians (linear in S 1 ) that include the spin part of (1.1), i.e. the second and third
terms, with only Jx 6= 0 and the periodic boundary condition (this boundary condition is not
essential). Matsui in [6, 7] enlarged a class of spin- 12 XZ-type models in which Gibbsian states
exist. The method was further developed by Datta and Kennedy in [8]. An application of
classical spins systems for constructing of quantum states was given in [9].
In [10] we showed how to find VΛ for a given Gibbsian ground state and established existence
of lro’s in S 1 and S 3 for a wide class of the spin- 12 XZ models (see also [11, 12]). This reference
contains the most simple proofs of the existence of lro’s in ground states of quantum many-body
systems. A reader may find a review of the results concerning several quantum orders in it (see
also [13, 14, 15]).
The ground state in (1.3) can be represented in the following equivalent form
X 1
X
0
e− 2 U∗ (sΛ ) ψΛ
ΨΛ (qΛ ) =
(sΛ )ψ0Λ (qΛ + ηφΛ (sΛ )),
U∗ = U 0 − µη 2
φ2x .
(1.5)
sΛ
x∈Λ
orthonormality
of the basis (see the beginning of the third section) and the
From (1.5),
R
R
equalities ψ02 (q)dq = 1, ψ02 (q)qdq = 0 it follows that
hˆ
qx qˆy iΛ = η 2 hφx (SΛ3 )φy (SΛ3 )iΛ = η 2 hφx (σΛ )φy (σΛ )i∗Λ ,
hSx3 Sy3 iΛ = hσx σy i∗Λ ,
(1.6)
where σx (sΛ ) = sx and
−1
hφx (σΛ )φy (σΛ )i∗Λ = Z∗Λ
X
sΛ
e−U∗ (sΛ ) φx (sΛ )φy (sΛ ),
Z∗Λ =
X
e−U∗ (sΛ ) .
sΛ
Equalities in (1.6) reduce a calculation of averages in our quantum systems to a calculation of
averages indexed by a star in Ising models. For a short-range ferromagnetic potential energy U∗
the first Griffiths inequality holds: hσ[A] i∗Λ ≥ 0. As a result the following statement (principle)
is true.
Proposition 1. Let J0 ≥ 0 in (1.2) and ferromagnetic lro occur in the Ising model with the
ferromagnetic potential energy U∗ given by (1.5), that is hσx σy i∗Λ > 0 uniformly in Λ, then
ferromagnetic lro occurs in oscillator variables and S 3 in the quantum spin-oscillator system
with the Hamiltonian (1.1) and ground state ΨΛ in (1.3).
Usefulness of Gibbsian ground states is explained by comparative simplicity of a proof of
existence of lro. Our results show that Gibbsian ground states are expected to appear in many
quantum spin-oscillator systems with non-trivial interactions.
Our paper is organized as follows. In Section 2 we formulate our main results in a lemma
and theorem. In next sections we prove them.
4
2
T.C. Dorlas and W.I. Skrypnik
Main result
We establish that Gibbsian ground states exist for JA depending on qΛ if
VΛ = −
X
1
3
JA e− 2 WA (SΛ ) ,
WA (SΛ3 ) = U (SΛ3A ; qΛ ) − U (SΛ3 ; qΛ ),
(2.1)
A⊆Λ
3 , S3
where SΛ3A = (−SA
Λ\A ). VΛ determines an unbounded operator. Negative JA generate
positive functions VΛ and this enables us to prove the following lemma.
Lemma 1. Let VΛ be given by (2.1) and have a domain D(VΛ ), JA be bounded negative functions
and U in (2.1) coincide with U in (1.4) Then HΛ is positive definite and essentially self-adjoint
on the set (⊗ C2 )|Λ| ⊗S(R|Λ| )∩D(VΛ ) that contains (⊗ C2 )|Λ| ⊗C0∞ (R|Λ| ) and ΨΛ , given by (1.3),
is an eigenfunction of the Hamiltonian (1.1) with the zero eigenvalue and is its (unique) ground
state if JA ≤ 0 (if the uniform bound Jx ≤ J− < 0 holds).
Remark 1. If the functions JA are only negative then ΨΛ is the ground state of the self-adjoint
extension of the Hamiltonian preserving positive definiteness.
The most simple translation invariant short-range U 0 is ferromagnetic
U 0 (sΛ ) = −α
X
JA0 s[A] ,
JA0 ≥ 0,
(2.2)
A⊆Λ
where JA0 = 0 for odd |A|, α ≥ 0.
Theorem 1. Let all the conditions of Lemma 1 be satisfied, J0 ≥ 0, J0 (x; A) = 0 for even |A|,
0 ≥ J¯ and U 0 be given by (2.2). Then
J0 (0; 1), J0,1
I. For a sufficiently large β = (η 2 µ+α)J¯ > 1 there exist ground state lro’s in S 3 and oscillator
variables for d ≥ 2;
II. Let the positive constants C, Bj , j = 1, 2, independent of Λ, exist such that |φx (sΛ )| ≤ C
(j)
WA (sΛ ) =
X
j
|φjx′ (sA
Λ ) − φx′ (sΛ )| ≤ Bj ,
j = 1, 2,
x′ ∈Λ
where |A| = 2. Then ground state ferromagnetic lro in S 1 occurs in arbitrary dimension d.
If U 0 = 0 then VΛ from (2.1) is given by
VΛ = −
X
JA v[A] ,
vx = cosh ux + Sx3 sinh ux ,
ux = 2ηµφx (qΛ ).
A⊆Λ
This equality follows from the equalities
WA (SΛ3 ) = −2
X
ux Sx3 ,
3
eaS = cosh a + S 3 sinh a,
(S 3 )2 = I .
x∈A
In the simplest case the conditions of item II of Theorem 1 can be checked without difficulty.
Proposition 2. Let φx be linear in S 3 in (1.2) and ||J0 ||1 =
P
x
|J0 (x)| < ∞, where the sum-
mation is performed over Zd . Then the conditions of item II of Theorem 1 are satisfied.
Three Order Parameters in Quantum XZ Spin-Oscillator Models
5
Note that if one uses the Pauli matrices with 12 instead of the unity as matrix elements
then VΛ should be changed by adding to WA the number −|A| ln 2.
If one chooses the anti-ferromagnetic U 0 in (2.1), specifically,
X
X
U 0 (sΛ ) = µη 2
φ2x (sΛ ) + α
sx sy ,
α > 0,
x∈Λ
hx,yi∈Λ
where x, y are nearest neighbors, then it can be easily proved that the spin lro in the third
direction will be anti-ferromagnetic, generating a staggered magnetization (spins at the even
and odd sublattices take different values).
The interesting and important property of the Hamiltonians with VΛ given by (2.1) is that
they are simply related to generators of stationary Markovian processes (see, also, [16]). We
believe that it is possible to apply the same mathematical technique for proving existence of
order and phase transitions in equilibrium quantum systems and non-equilibrium stochastic
systems (see [17]).
3
Proof of Lemma 1
For our purpose it is convenient to pass to a new representation. It is determined by the Hilbert
space of sequences of functions F (qΛ ; sΛ ), sx = ±1, which are found in the expansion of the
vector F ∈ L2Λ mentioned at the beginning of the introduction, with the scalar product
XZ
F1 (qΛ ; sΛ )F2 (qΛ ; sΛ )dqΛ ,
(F1 , F2 ) =
sΛ
(F1 (qΛ ), F2 (qΛ ))0 =
X
F1 (qΛ ; sΛ )F2 (qΛ ; sΛ ),
sΛ
where the integration is performed over R|Λ| . Here we took into account the orthonormality of
the basis, i.e. the equality
Y
δsx ,s′x ,
(Ψ0Λ (sΛ ), Ψ0Λ (s′Λ ))0 = δ(sΛ ; s′Λ ) =
x∈Λ
where δs,s′ is the Kronecker delta. Let
hx = −∂x2 + µ2 (qx + ηφx (SΛ3 ))2 − µ,
h0x = −∂x2 + µ2 qx2 − µ = (−∂x + µqx )(∂x + µqx )
and
hΛ =
X
hx ,
h0Λ =
X
h0x .
x∈Λ
x∈Λ
(Tx (φ)F )(qΛ ) =
X
0
F (qx + ηφx (sΛ ), qΛ\x ; sΛ )ψΛ
(sΛ ),
sΛ
then
hx = Tx (φ)h0x Tx−1 (φ),
hΛ = TΛ h0Λ TΛ−1 .
Here we took into account that differentiation commutes with TΛ .
Our Hamiltonian is rewritten as follows
X
1
3
1
HΛ = hΛ +
JA P[A] ,
PA = S[A]
− e− 2 WA (SΛ ) .
A⊆Λ
6
T.C. Dorlas and W.I. Skrypnik
The remarkable fact is that the symmetric operator PA and the harmonic operator hx have both
common eigenvector ΨΛ with the zero eigenvalue (see Remark 3 in the end of the paper). Note
that the space of ground states of the operator hx (eigenfunctions with the zero eigenvalue) is
2|Λ| -fold degenerate since Sx3 is diagonal and the Laplacian is translation invariant. From (1.4)
and the definition of Tx (φ) if follows that Tx−1 (φ)ΨΛ is equal to ψ0 (qx ) multiplied by a function
independent of qx
X 1
0
(sΛ )ψ0Λ (qΛ\x + ηφΛ\x (sΛ\x ))ψ0 (qx ).
e− 2 U∗ (sΛ ) ψΛ
Tx−1 (φ)ΨΛ =
sΛ
Hence h0x Tx−1 (φ)ΨΛ = 0 and hx ΨΛ = Tx (φ)h0x Tx−1 (φ)ΨΛ = 0. The proof that ΨΛ is an eigenvector with the zero eigenvalue of PA is inspired by our previous paper [10]. For simplicity we will
omit qΛ in the expression for U in (1.3). Taking into consideration the equalities
0
0
(sΛ ) = sx ψΛ
(sΛ ),
Sx3 ψΛ
0
0 A
1
0
(sΛ\A , −sA ),
S[A]
ψΛ
(sΛ ) = ψΛ
(sΛ ) = ψΛ
we obtain
(ψ0Λ )−1 PA ΨΛ =
X
1
1
0
0
(ψΛ
(sΛ\A , −sA ) − e− 2 WA (sΛ ) ψΛ
(sΛ ))e− 2 U (sΛ )
sΛ
=
X
1
1
A
0
0
(sΛ\A , −sA )e− 2 U (sΛ ) − ψΛ
(ψΛ
(sΛ )e− 2 U (sΛ ) )
sΛ
=
X
1
A
1
A
0
(sΛ ) = 0.
(e− 2 U (sΛ ) − e− 2 U (sΛ ) )ψΛ
sΛ
Here we changed signs of the spin variables sA in the first term in the sum in sΛ .
Positive definiteness of the Hamiltonian follows from the following proposition.
Proposition 3. The operator −PA is positive definite on D(VΛ ).
Proof . VΛ is an operator of multiplication by infinite differentiable functions on each onedimensional spin subspace and its domain contains (⊗C2 )|Λ| ⊗ C0∞ (R|Λ| ).PThis domain coincides
|qx ||φx |.
with the direct sum of 2|Λ| copies of of L2 (R|Λ| , e|q|0 dqΛ ), where |q|0 =
x∈Λ
R
The scalar product in the Hilbert space L2Λ is given by (F1 , F2 ) = (F1 (qΛ ), F2 (qΛ )) dqΛ ,
where the integration is performed over R|Λ| . We have to show that −(PA F (qΛ ), F (qΛ ))0 ≥ 0.
Let us define the operator
1
1
3
3
PA+ = e 2 U (SΛ ) PA e− 2 U (SΛ ) .
0 that
It is not difficult to check on the basis ψΛ
1
3
1
PA+ = e− 2 WA (SΛ ) (S[A]
− I),
where I is the unit operator. Here we used the following equality
1
1
3A
3
1
1 − 2 U (SΛ )
= S[A]
e
.
e− 2 U (SΛ ) S[A]
For the operator PA+ we have
X
0
(PA+ F )(qΛ ; sΛ )ψΛ
(sΛ )
PA+ F =
sΛ
and
1
(PA+ F )(qΛ ; sΛ ) = −e− 2 WA (sΛ ) (F (qΛ ; sΛ ) − F (qΛ ; sA
Λ )).
Three Order Parameters in Quantum XZ Spin-Oscillator Models
7
It is convenient to introduce the new scalar product
XZ
3)
−U (SΛ
F1 (qΛ , sΛ )F2 (qΛ , sΛ )e−U (sΛ ) dqΛ
(F1 , F2 )U = (e
F1 , F 2 ) =
sΛ
=
Z
(F1 (qΛ ), F2 (qΛ ))0U dqΛ
=
Z
3
(e−U (SΛ ) F1 (qΛ ), F2 (qΛ ))0 dqΛ .
The operator PA+ is symmetric with respect to the new scalar product since
1
1
3
3
(PA+ F1 (qΛ ), F2 (qΛ ))0U = (PA e− 2 U (SΛ ) F1 (qΛ ), e− 2 U (SΛ ) F2 (qΛ ))0 .
(3.1)
It is not difficult to check that
X 1
A
e− 2 [U (sΛ )+U (sΛ )] (F (qΛ ; sΛ ) − F (qΛ ; sA
−(PA+ F (qΛ ), F (qΛ ))0U =
Λ ))F (qΛ ; sΛ )
sΛ
1 X − 1 [U (sΛ )+U (sA )]
Λ (F (q ; s ) − F (q ; sA ))2 ≥ 0.
=
e 2
Λ Λ
Λ Λ
2 s
(3.2)
Λ
Here we took into account that the function under the sign of exponent is invariant under
changing signs of the spin variables sA . From (3.1), (3.2) it follows that −PA ≥ 0.
P
JA PA is positive definite since for the negaFrom Proposition 3 it follows that HΛ1 =
A⊆Λ
tive JA we have
(HΛ1 F, F )
=
XZ
JA (qΛ )(PA F (qΛ ), F (qΛ ))0 dqΛ ≥ 0.
A⊆Λ
The fact that the Hamiltonian is essentially self-adjoint is derived from the following proposition (see example X.9.3 in [18]).
Proposition 4. The operator −∆ + µ2 x2 + V + (y, x), where (y, x) is the Euclidean scalar
n
P
∂
product of x = xj , j = 1, . . . , n with the constant vector y, ∆ =
∂xj , µ 6= 0 if y 6= 0, V is
j=1
2
the operator of multiplication by a non-negative function V (x) ∈ L2 (Rn , e−x dx), x2 = (x, x),
is essentially self-adjoint on S(Rn ) ∩ D(V ), i.e. the intersection of the Schwartz space and the
domain of V .
Proof . The following two inequalities are valid
µ2 ||x2 ψ|| ≤ ||(−∆ + V + µ2 x2 )ψ|| + 2µ2 n||ψ||,
(y, x) ≤ ǫx2 +
n 2
|y| ,
4ǫ
(3.3)
where µ is a real number, |y| = max |y|j , || · || is the scalar product in the Hilbert space of
j
square integrable functions. We tacitly assume that x2 means the operator of multiplication of
a squared variable. The first inequality follows from the inequalities (∂j = ∂x∂ j )
(−∆ + V + µ2 x2 )2 ≥ (µ2 x2 )2 − µ2 (∆x2 + x2 ∆),
−(∂j2 x2k + x2k ∂j2 ) = −(2xk ∂j2 xk + 2δj,k ) ≥ −2δj,k ,
n
X
−(∆x2 + x2 ∆) ≥ −
(∂j2 x2j + x2j ∂j2 ).
j=1
8
T.C. Dorlas and W.I. Skrypnik
Here we took into account that V is positive and [∂j , xk ] = ∂j xk − xk ∂j = δj,k . As a result
|y|2
2 2
−2
||(y, x)ψ|| ≤ a||(−∆ + V + µ x )ψ|| + b||ψ||,
a = µ ǫ,
b=n 2+
.
4ǫ
For µ 6= 0 number a can be arbitrary small and from the Kato–Rellich theorem [18] it follows
that the essential domain of −∆ + V + µ2 x2 + (y, x) coincides with the essential domain of
−∆+V +µ2 x2 . But the last one coincides with S(Rn )∩D(V ) [18, Theorem X.59]. Inequality (3.3)
is sufficient, also, for the proof of the proposition in the case µ = y = 0 (this is explained in
Example X.3 in [18]).
Since SΛ3 is a diagonal operator on (⊗C2 )|Λ| the operator
X
X
X
qx φx (SΛ3 )
qx2 + VΛ + 2ηµ2
∂x2 + µ2
−
x∈Λ
x∈Λ
x∈Λ
is the direct sum of 2|Λ| copies of the minus |Λ|-dimensional Laplacian plus the three functions
coinciding with the three functions in Proposition 4. From Proposition 4 it follows that this
2 |Λ| ⊗
operator is essentially self-adjoint on the set (⊗C2 )|Λ| ⊗ S(R|Λ| ) ∩ D(VP
Λ ) that contains (⊗C )
2
3
∞
|Λ|
2
φx (SΛ ) and the operator
C0 (R ). The same is true for the Hamiltonian since operator (ηµ)
x∈Λ
depending on SΛ1 in its expression are bounded.
The operators hΛ , HΛ1 are positive definite on the dense set (⊗C2 )|Λ| ⊗ S(R|Λ| ) ∩ D(VΛ ) which
is the essential set for HΛ . This implies that HΛ is positive definite on its domain D(HΛ ) and ΨΛ
is its ground state.
Proof of uniqueness. We have to establish that the symmetric semigroup PΛt , generated
by −HΛ , maps non-negative functions into (strictly) positive functions (increases positivity) and
this will imply that the ground state is unique (see Theorem XIII.44 in [18]). We will establish
this property with the help of a perturbation expansion. The kernel of the semigroup P1t ,
generated by hΛ + VΛ , is expressed in terms of the Feynman–Kac (FK) formula [18, 19]
Z
Zt
′
, s′Λ ) = δsΛ ,s′Λ PqtΛ ,q′ (dwΛ ) exp − VΛ+ (wΛ (τ ), sΛ )dτ ,
P1t (qΛ , sΛ ; qΛ
Λ
0
where VΛ+ (qΛ ; sΛ ) = VΛ (qΛ ; sΛ )+
P
[µ2 (qx +ηφx (sΛ ))2 −µ], PqtΛ ,q′ (dwΛ ) =
Λ
x∈Λ
Q
x∈Λ
Pqtx ,qx′ (dwx ) is the
conditional Wiener measure and wΛ (t) is the sequence of continuous paths. The semigroup P t
is represented as a perturbation series in powers of V0
X
1
V0 = −
JA S[A]
.
A⊆Λ
This series is convergent in the uniform operator norm [20] since V0 is a bounded operator. Its
perturbation expansion is given by
t
P =
X
n≥0
Pnt ,
Pnt
=
Z
0≤τ1 ≤τ2 ≤···≤τn ≤t
dτ1 · · · dτn P1τ1
n+1
Y
j=2
where τn+1 = t . We now use the following simple inequality
Z t
|VΛ+ (wΛ (τ ), sΛ )|dτ
0
τ −τj−1
(V0 P1 j
),
Three Order Parameters in Quantum XZ Spin-Oscillator Models
≤ V¯ (wΛ ) = |J|Λ
Z t"
(
¯0 + 2ηµ
exp αU
0
X
)
φ¯x |wx (τ )|
x∈Λ
+
X
9
#
(µ2 (|wx (τ )| + η φ¯x )2 − µ) dτ,
x∈Λ
¯0 = max U0 (sΛ ), |J|Λ = sup P |JA | , φ¯x = max |φx (sΛ )|. Let
where U
sΛ
sΛ
qΛ A⊆Λ
P¯1t (qΛ ; qΛ ) =
Z
¯
PqtΛ ,q′ (dwΛ )e−V (wΛ ) ,
V− = −J−
Λ
X
Sx1
x∈Λ
then it follows from the positivity of the the kernel P1t and V0 that
′
P t (qΛ , sΛ ; qΛ
, s′Λ ) ≥ P¯1t (qΛ ; qΛ )
X tn
n≥0
n!
V−n (sΛ ; s′Λ ).
(3.4)
Here we utilized the semigroup property of P¯ t , the inequalities ea ≥ e−|a| , V0 ≥ V− ,
τ −τj−1
(V0 P1 j
′
′
)V− (sΛ ; , s′Λ ).
, s′Λ ) ≥ P¯ τj −τj−1 (qΛ ; qΛ
)(qΛ , sΛ ; qΛ
Now, it is easily proved as in [10] that the matrix V− is irreducible. As a result there exists
a positive integer n such that and that V−n , has positive non-diagonal elements [21, 22]. Hence
the kernel in the left-hand side of (3.4) is positive.
4
Order parameters
In this section we will prove Theorem 1, i.e. occurrence of different types of lro in the considered
quantum systems. In the proof we will rely on the following basic theorem.
Theorem 2. Let the ferromagnetic short-range potential energy
P U of the classical Ising model
on the hyper-cubic lattice Zd with the partition function ZΛ = exp{−βU (sΛ )} be given by
sΛ
U (sΛ ) = −
X
(4.1)
ϕ(A)s[A] ,
A⊆Λ
where ϕ(A) ≥ 0, sx = ±1. Let, also, the uniform bound ϕ(x, y) ≥ ϕ¯ > 0 for nearest neighbors
x, y hold and ϕ(A) = 0 for A with odd number of sites. Then for a sufficiently large ϕβ
¯ >1
and the dimension d ≥ 2 there is the ferromagnetic lro, that is, for the Gibbsian two point spin
average the uniform in Λ bound holds
hσx σy iΛ > 0,
where σx (sΛ ) = sx and the magnetization (an order parameter) MΛ = |Λ|−1
P
sx is non-zero
x∈Λ
in the thermodynamic limit Λ → Zd .
The proof of this theorem is based on an application of the generalized Peierls principle
(argument). It will be given in the end of this section (see, also, [17, 26]. The next theorem is
the consequence of the basic theorem.
Proof of item I of Theorem 1. Condition (1.2) shows that
X
x∈Λ
φ2x (sΛ ) = |Λ| + 2
X
A⊆Λ
J2 (A)s[A] +
X
x∈Λ
X
x6∈A⊆Λ
2
J0 (x; A)s[A] .
10
T.C. Dorlas and W.I. Skrypnik
where J2 (x ∪ A) = J0 (x; A) and J2 (A) = 0 for odd |A|. The last term is equal to
X X
J02 (x; A) + 2
X
X
J0 (x; A1 )J0 (x; A2 )s[A1 ∆A2 ] ,
x∈Λ x6∈A1 ,A2 ⊆Λ
x∈Λ x6∈A⊆Λ
where s∅ = 1 and A1 ∆A2 = (A1 ∪
PA2 )\(A1 ∩ A2 ). Due to translation invariance of interaction
the first term is bounded by |Λ| J02 (0; A), where the summation is performed over Zd , and
A
this expression is finite since the interaction is short-range. Hence subtracting a finite constant
proportional to |Λ| from U∗ one sees that the result admits representation (4.1) with positive J∗
instead of ϕ such that J∗ (A) = 0 for odd |A| and for nearest neighbors x, y the inequality
¯ The basic theorem and Proposition 1 imply occurrence of ferromagnetic
J∗ (x, y) ≥ (2η 2 µ + α)J.
3
lro in S and oscillator lro.
0 (s )ψ (q ). Then (1.2) and the defiProof of item II of Theorem 1. Let ψΛ (qΛ ; sΛ ) = ψΛ
Λ 0Λ Λ
1
nition of S imply that
X
Sx1 Sy1 ΨΛ (qΛ ) =
1
e− 2 U (sΛ ;qΛ ) ψΛ (qΛ ; sx,y
Λ )=
sΛ
X
x,y
1
e− 2 U (sΛ
;qΛ )
ψΛ (qΛ ; sΛ ).
sΛ
Taking into account also the orthonormality of the basis one obtains
XZ
x,y
1
2
hSx1 Sy1 iΛ = ZΛ−1
e− 2 [U (sΛ ;qΛ )+U (sΛ ;qΛ )] ψ0Λ
(qΛ )dqΛ
sΛ
=
ZΛ−1
X
e
η2 µ
4
P
x′ ∈Λ
2
(φx′ (sΛ )+φx′ (sx,y
Λ ))
α
x,y
e− 2 [U0 (sΛ
)+U0 (sΛ )]
sΛ
−α
B
2 0
≥e
ZΛ−1
X
e
η2 µ
4
P
x′ ∈Λ
2
(φx′ (sΛ )+φx′ (sx,y
Λ ))
e−αU0 (sΛ ) .
sΛ
We also have
X
2
(φx′ (sΛ ) + φx′ (sx,y
Λ ))
x′ ∈Λ
=
X
[3φ2x′ (sΛ ) + φ2x′ (sx,y
Λ )] + 2
x′ ∈Λ
x′ ∈Λ
=
X
4φ2x′ (sΛ ) +
x′ ∈Λ
≥4
φx′ (sΛ )[−φx′ (sΛ ) + φx′ (sx,y
Λ )]
X
X
[−φ2x′ (sΛ ) + φ2x′ (sx,y
Λ )] + 2
x′ ∈Λ
X
X
φx′ (sΛ )[−φx′ (sΛ ) + φx′ (sx,y
Λ )]
x′ ∈Λ
φ2x′ (sΛ ) − B2 − 2CB1 .
x′ ∈Λ
This yields
hSx1 Sy1 iΛ
2
B
− η 4µ (B2 +2B1 ) − α
2 0
≥e
e
ZΛ−1
X
η2 µ
e
P
x′ ∈Λ
φ2x′ (sΛ )
e−αU0 (sΛ )
sΛ
2
− η 4µ (B2 +2CB1 )
=e
−α
B
2 0
e
.
Proof of Proposition 2. It is obvious that |φx (sΛ )| ≤ ||J0 ||1 and that
′
′
′
′
|φx′ (sx,y
Λ ) − φx′ (sΛ )| = | − 2[sy J0 (y − x ) + sx J0 (x − x )]| ≤ 2[|J0 (y − x )| + |J0 (x − x )|].
Three Order Parameters in Quantum XZ Spin-Oscillator Models
11
(1)
As a result Wx,y (sΛ ) ≤ 4||J0 ||1 . Further
"
#2
X
x,y
2
2
′
′
′
|φx′ (sΛ ) − φx′ (sΛ )| =
J0 (z − x )sz − sy J0 (y − x ) − sx J0 (x − x )
z∈Λ\(x,y)
"
#2
X
−
J0 (z − x′ )sz + sy J0 (y − x′ ) + sx J0 (x − x′ )
z∈Λ\(x,y)
X
= −4
J0 (z − x′ )sz (sy J0 (y − x′ ) + sx J0 (x − x′ ))
z∈Λ\(x,y)
X
|J0 (z − x′ )|(|J0 (y − x′ )| + |J0 (x − x′ )|) ≤ 4||J0 ||1 (|J0 (y − x′ )| + |J0 (x − x′ )|).
≤4
z∈Λ
(2)
Hence Wx,y (sΛ ) ≤ 8||J0 ||21 .
1
Proof of Theorem 2. Let χ±
x = 2 (1 ± σx ) then one obtains
−
4hχ+
x χy iΛ = 1 + hσx iΛ − hσy iΛ − hσx σy iΛ .
Since the systems are invariant under the transformation of changing signs of spins the third
and the second terms in the right-hand side of last equality are equal to zero and
−
hσx σy iΛ = 1 − 4hχ+
x χy iΛ .
Hence if
−
hχ+
x χy iΛ <
1
4
(4.2)
then the ferromagnetic lro occurs, i.e.
hσx σy iΛ ≥ a > 0,
(4.3)
where a is independent of Λ. If one succeeds in proving that there exists a positive function E0 (β)
and positive constants a, a′ independent of Λ such that
′ E0 (β)
−
hχ+
x χy iΛ ≤ a e
(4.4)
and proves that E0 is increasing at infinity then (4.2), (4.3) will hold for a sufficiently large
inverse temperature β. The Peierls principle reduces the derivation of (4.4) to the derivation of
the contour bound.
Peierls principle. Let the contour bound hold
h
Y
−
−|Γ|E
,
χ+
x χy iΛ ≤ e
(4.5)
hx,yi∈Γ
where h·iΛ denotes the Gibbs average for the spin system confined to a compact domain Λ,
Γ is a set of the nearest neighbors, adjacent to the (connected) contour, i.e. a boundary of the
connected set of unit hypercubes centered at lattice sites. Then (4.4) is valid with E0 = a′′ E,
where a′′ is a positive constant independent of Λ.
12
T.C. Dorlas and W.I. Skrypnik
Proof of contour bound. Bricmont and Fontain derived the contour bound for the spin systems with the potential energy (4.1) with the help of the second Griffiths [23] and Jensen
inequalities [24] (see also [25, 26])
Z
Z
f
hσ[A] σ[B] iΛ[Γ] − hσ[A] iΛ[Γ] hσ[B] iΛ[Γ] ≥ 0,
e dµ ≥ exp
f dµ ,
where dµ is a probability measure on a measurable space. Their proof starts form the inequality
β
β
β
β
− 2 σx σ y 2 σx σ y + −
−
− 2 σ x σy
−
χ+
e
χx χy ≤ e− 2 σx σy χ+
.
x χy = e
x χy ≤ e
As a result (β ′ = ϕβ)
¯
′
h
Y
−
χ+
x χy iΛ
− β2
≤ he
P
σx σy
iΛ = he
hx,yi∈Γ
β′
2
P
hx,yi∈Γ
σ x σy
i−1
Λ[Γ]
hx,yi∈Γ
′
− β2
≤e
P
hσx σy iΛ[Γ]
hx,yi∈Γ
= e−EΓ ,
where h·, ·iΛ[Γ] is the average corresponding to the potential energy
UΓ (qΛ ) = U (sΛ ) +
ϕ¯ X
sx sy .
2
hx,yi∈Γ
In the last line we applied the Jensen inequality. From the second Griffiths inequality it follows
that the average hσx σy iΛ[Γ] is a monotone increasing function in ϕA . So, in the potential energy
determining this average we can put ϕA = 0, except A = hx, yi and leave the coefficient ϕ¯ in
front of the bilinear nearest-neighbor pair potential in (4.1) without increasing the average. This
leads to
′ X
X
β′
β′
−1 β
′
Z2 (β) =
e 2 s1 s2 .
hσx σy iΛ[Γ] ≥ hσσ i = Z2
s1 s2 e 2 s1 s2 ,
2
s1 ,s2 =±1
s1 ,s2 =±1
That is,
E = 2−1 β ′ hσσ ′ i
EΓ ≥ |Γ|E,
or
E = β(e2
−1 β ′
− e−2
−1 β ′
)(e2
−1 β ′
+ e−2
−1 β ′
Here we used in the denominator the inequality e−2
if β ′ tends to infinity. This implies (4.5).
5
′
)−1 ≥ 2−1 β ′ (1 − e−β ).
−1 β ′
≤ e2
−1 β ′
. Obviously, E tends to infinity
Discussion
We showed that in the considered lattice spin-boson models with JA ≤ 0 ground states are
Gibbsian and the ground state averages for special observables are reduced to averages in classical
Ising models. This means that existence of ground states order parameters is connected with
existence of order parameters in the associated Ising models and that a breakdown of symmetries
in the quantum systems is determined by a breakdown of symmetries in Ising models. We
considered the free boundary conditions implying that for the cases of the perturbation VΛ ,
Three Order Parameters in Quantum XZ Spin-Oscillator Models
13
considered in the two theorems, the ground state averages of qx , Sx3 are zero, that is hˆ
qx iΛ = 0,
hSx3 iΛ = 0 if the associated Ising potential energy is an even function. In order to make such the
averages non-zero (explicit symmetry breaking) one has to introduce special boundary conditions
(quasi-averages) which have to single out pure Gibbsian states in the associated Ising models. It
is known [28] that for the two-dimensional ferromagnetic Ising nearest-neighbor model there are
two boundary conditions which generate pure states and that every other state is a convex linear
combination of these two states. A discussion of a construction of ground states in lattice spin
and fermion quantum systems with an explicit symmetry breaking a reader may find in [29].
Remark 2. Translation invariance means that
Jx1 ,...,xn = J0,x2 −x1 ,...,xn −x1 ,
J0 (x; x1 , . . . , xn ) = J0 (0; x1 − x, . . . , xn − x)
where J, J0 are symmetric functions. The short-range character of interaction means that
X
X
|J0 (x; A)| < ∞.
|Jx,A | < ∞,
max
max
x
A
x
A
Remark 3. If only one-point sets are left in the sum for VΛ then the expression for HΛ can be
rewritten in the following way
X
HΛ =
Hx .
x∈Λ
The property of the ground state ΨΛ to be a ground state with the zero eigenvalue of a local
Hamiltonian Hx was found earlier for special isotropic anti-ferromagnetic Heisenberg chains with
valence bond ground state in [27].
References
[1] Kobayashi K., Dynamical theory of the phase transition in KH2 PO4 -type ferroelectric crystals, J. Phys.
Soc. Japan 24 (1968), 497–508.
[2] Villain J., Stamenkovic S., Atomic motion in hydrogen-bond ferroelectrics, Phys. Stat. Sol. 15 (1966), 585–
596.
[3] Kurbatov A., Plechko V., Exactly solvable model with order-disorder phase transition for ferroelectrics,
Teoret. Mat. Fiz. 26 (1976), 109–116.
[4] Bogolyubov N.N. Jr., Brankov J.G., Kurbatov A.M., Tonchev N.C., Zagrebnov V.A., Method of approximating Hamiltonian in statistical physics, Sophia, 1981.
[5] Kirkwood J., Thomas L., Expansions and phase transitions for the ground state of quantum Ising lattice
systems, Comm. Math. Phys. 88 (1983), 569–580.
[6] Matsui T., A link between quantum and classical Potts models, J. Statist. Phys. 59 (1990), 781–798.
[7] Matsui T., Uniqueness of the translationally invariant ground state in quantum spin systems, Comm. Math.
Phys. 126 (1990), 453–467.
[8] Datta N., Kennedy T., Expansions of one quasiparticle states in spin-1/2 systems, J. Statist. Phys. 108
(2002), 373–399, cond-mat/0104199.
[9] Goderis D., Maes C., Constructing quantum dissipations and their reversible states from classical interacting
spin systems, Ann. Inst. H. Poincar´e Phys. Th´eor. 55 (1991), 805–828.
[10] Dorlas T., Skrypnik W., On quantum XZ 1/2 spin models with Gibbsian ground states, J. Phys. A: Math.
Gen. 37 (2004), 66623–6632.
[11] Skrypnik W., Order parameters in XXZ-type spin
2 (2006), 011, 6 pages, math-ph/0601060.
1
2
quantum models with Gibbsian ground states, SIGMA
[12] Skrypnik W., Quantum spin XXZ-type models with Gibbsian ground states and double long-range order,
Phys. Lett. A 371 (2007), 363–373.
14
T.C. Dorlas and W.I. Skrypnik
[13] Messager A., Nachtergaele B., A model with simultaneous first and second order phase transitions, J. Statist.
Phys. 122 (2006), 1–14, cond-mat/0501229.
[14] van Enter A.C.D., Shlosman S.B., First-order transitions for n-vector models in two and more dimensions;
rigorous proof, Phys. Rev. Lett. 89 (2002), 285702, 3 pages, cond-mat/0205455.
[15] van Enter A.C.D., Shlosman S.B., Provable first-order transitions for liquid crystal and lattice gauge models
with continuous symmetries, Comm. Math. Phys. 255 (2005), 21–32, cond-mat/0306362.
[16] Thomas L.E., Quantum Heisenberg ferromagnets and stochastic exclusion processes, J. Math. Phys. 21
(1980), 1921–1924.
[17] Skrypnik W., Long-range order in nonequilibrium systems of interacting Brownian linear oscillators,
J. Statist. Phys. 111 (2003), 291–321.
[18] Reed M., Simon B., Methods of modern mathematical physics, Vols. II, IV, Academic Press, 1975.
[19] Ginibre J., Reduced density matrices of quantum gases. I. Limit of infinite volume, J. Math. Phys. 6 (1964),
238–251.
[20] Kato T., Perturbation theory for linear operators, Springer-Verlag, 1966.
[21] Gantmacher F.R., Matrix theory, 4th ed., Nauka, Moscow, 1988.
[22] Gantmacher F.R., Applications of the theory of matrices, Interscience Publ., New York, 1959.
[23] Kelly D., Sherman S., Inequalities on correlations in Ising ferromagnets, J. Math. Phys. 9 (1968), 466–483.
[24] Bricmont J., Fontaine J.-R., Correlation inequalities and contour estimates, J. Statist. Phys. 26 (1981),
745–753.
[25] Fr¨
ohlich J., Lieb E., Phase transitions in anisotropic lattice spin systems, Comm. Math. Phys. 60 (1978),
233–267.
[26] Skrypnik W., Long-range order in linear ferromagnetic oscillator systems. Strong pair quadratic n-n potential, Ukrainian Math. J. 56 (2004), 964–972.
[27] Affleck I., Kennedy T., Lieb E., Tasaki H., Valence bond ground states in isotropic quantum antiferromagnets, Comm. Math. Phys. 115 (1988), 477–528.
[28] Izenman M., Translation invariance and instability of phase coexistence in the teo dimensional Ising system,
Comm. Math. Phys. 73 (1980), 84–92.
[29] Koma T., Tasaki H., Symmetry breaking and finite-size effects in quantum many-body systems, J. Statist.
Phys. 76 (1994), 745–803, cond-mat/9708132.
SIGMA 4 (2008), 007, 14 pages
Three Order Parameters in Quantum XZ
Spin-Oscillator Models with Gibbsian Ground States⋆
Teunis C. DORLAS
†
and Wolodymyr I. SKRYPNIK
‡
†
Dublin Institute for Advanced Studies, School of Theoretical Physics,
10 Burlington Road, Dublin 4, Ireland
E-mail: dorlas@stp.dias.ie
‡
Institute of Mathematics of NAS of Ukraine, Kyiv, Ukraine
E-mail: skrypnik@imath.kiev.ua
Received October 29, 2007, in final form January 08, 2008; Published online January 17, 2008
Original article is available at http://www.emis.de/journals/SIGMA/2008/007/
Abstract. Quantum models on the hyper-cubic d-dimensional lattice of spin- 21 particles
interacting with linear oscillators are shown to have three ferromagnetic ground state order
parameters. Two order parameters coincide with the magnetization in the first and third
directions and the third one is a magnetization in a continuous oscillator variable. The
proofs use a generalized Peierls argument and two Griffiths inequalities. The class of spinoscillator Hamiltonians considered manifest maximal ordering in their ground states. The
models have relevance for hydrogen-bond ferroelectrics. The simplest of these is proven to
have a unique Gibbsian ground state.
Key words: order parameters; spin-boson model; Gibbsian ground state
2000 Mathematics Subject Classification: 82B10; 82B20; 82B26
1
Introduction
In this paper we consider quantum lattice models of oscillators interacting with spins whose variables are indexed by the sites of a hyper-cube Λ with the finite number of sites |Λ| in the
hyper-cubic lattice Zd . Interaction is considered to be short-range and translation invariant.
The corresponding Hamiltonian HΛ is expressed in terms of the oscillators variables qΛ =
(qx , x ∈ Λ) ∈ R|Λ| and spin 12 Pauli matrices SΛl = (Sxl , x ∈ Λ, l = 1, 3), defined in the tensor product of the 2|Λ| -dimensional Euclidean space and the space of square integrable functions
L2Λ = (⊗E2 )|Λ| ⊗ L2 (R|Λ| ),
HΛ =
X
x∈Λ
X
1
JA S[A]
− ∂x2 + µ2 (qx + ηφx (SΛ3 ))2 − µ +
+ VΛ ,
µ ≥ 0,
η ∈ R,
(1.1)
A⊆Λ
where ∂x is the partial derivative in qx , JA and VΛ are real-valued measurable functions, the
first of which depends on qΛ and the second on SΛ3 , qΛ and the translation invariant φx is given
by (see Remark 2 in the end of the paper)
φx (sΛ ) =
X
J0 (x; A)s[A] ,
J0 (x; x) = 1.
(1.2)
A⊆Λ
⋆
This paper is a contribution to the Proceedings of the Seventh International Conference “Symmetry in
Nonlinear Mathematical Physics” (June 24–30, 2007, Kyiv, Ukraine). The full collection is available at
http://www.emis.de/journals/SIGMA/symmetry2007.html
2
T.C. Dorlas and W.I. Skrypnik
For products of operators, functions and variables we use the following notation: B[A] =
Q
Bx .
x∈A
The scalar product in (⊗E2 )|Λ| , L2Λ will be denoted by (·, ·)0 , (·, ·), respectively. The Schwartz
space of test functions on Rn will be denoted by S(Rn ).
We require that the Hamiltonian is well defined and bounded from below on (⊗E2 )|Λ| ⊗
C0∞ (R|Λ| ), i.e. the tensor product of the 2|Λ| -dimensional Euclidean space and the space of
infinitely differentiable functions with compact supports. The ground state average for an
observable (operator) F is determined by
hF iΛ = ZΛ−1 (ΨΛ , F ΨΛ ),
ZΛ = (ΨΛ , ΨΛ ) = ||ΨΛ ||2 ,
where ZΛ is a partition function. For partial cases of F we have
hˆ
q[A] φ[A′ ] (SΛ3 )iΛ
=
ZΛ−1
Z
q[A] (ΨΛ (qΛ ), φ[A′ ] (SΛ3 )ΨΛ (qΛ ))0 dqΛ ,
where the integration is performed over R|Λ| and qˆ[A] is the operator of multiplication by qx .
0 (s ) of the Euclidean space (⊗E2 )|Λ| , diagonaliWe will employ the orthonormal basis ψΛ
Λ
3
0 (s ) = ⊗
0
0
zing SΛ , which is chosen in the following way: ψΛ
Λ
x∈Λ ψ (sx ), sx = ±1, ψ (1) = (1, 0),
0
1
0
0
3
0
0
2
ψ (−1) = (0, 1), S ψ (s) = ψ (−s), S ψ (s) = sψ (s). For F ∈ LΛ we have the following
decomposition
F (qΛ ) =
X
0
(sΛ ),
F (qΛ ; sΛ )ψΛ
sΛ
where the summation is performed over the |Λ|-fold Cartesian product (−1, 1)|Λ| of the set (−1, 1).
The Hamiltonians in (1.1) are employed in hydrogen-bond ferroelectric crystal models, considered in [1, 2, 3], and describe interaction between heavy ions (oscillators with constant frequency) and protons (spins). The second term with JA = 0, |A| ≥ 2 corresponds to the energy
of protons, tunneling along hydrogen bonds from one well to another, and Jx is associated
with
P
3
J1 (A)S[A]
the tunneling frequency. The last term in the expression for our Hamiltonian VA =
A⊆Λ
describes many-body interaction between protons (J1 (A) is the intensity of the |A|-body interaction).
A rigorous analysis of a mean-field version of the
Hamiltonian in (1.1) with φx (SΛ3 ) given
P
in (1.2) that is linear in S 3 , Jx 6= 0, VΛ = −(µη)2
φ2x (SΛ3 ), JA = 0 for |A| ≥ 2 and J0 (x; y)
x∈Λ
uniformly in lattice sites proportional to |Λ|−1 was carried out in [3] in the framework of the Bogolyubov approximating Hamiltonian method [4] and occurrence of spin and oscillator orderings
(the corresponding order parameters are non-zero) for two-body interaction between protons at
non-zero temperatures was proved. To establish such orderings for ground states without the
mean-field limit in a general case is an important task for a theory.
The oscillator and spin orderings are established if one proves the existence of ferromagnetic
oscillator and spin long-range orders (lro’s) in ground states for the corresponding Hamiltonians.
This means that the ground state averages hˆ
qx qˆy iΛ , hSxj Syj iΛ , j = 1, 3 are bounded uniformly in Λ
from below by positive numbers . Occurrence of the ferromagnetic lro’s implies the existence
P of
l
−1
Sxl ,
the spin order parameters (magnetizations in the first and third directions) MΛ = |Λ|
x∈Λ
P
qx in the thermodynamic limit
l = 1, 3, and the oscillator order parameter QΛ = |Λ|−1
x∈Λ
(Λ → Zd ) since the ground state averages of their squares are uniformly bounded in Λ from
below by a positive number.
Three Order Parameters in Quantum XZ Spin-Oscillator Models
3
In this paper we find functions VΛ , depending on oscillator variables, for which ground states
or eigenstates ΨΛ of the Hamiltonians in (1.1) are Gibbsian
X 1
0
e− 2 U (sΛ ;qΛ ) ψΛ
ΨΛ (qΛ ) =
(sΛ )ψ0Λ (qΛ ),
(1.3)
sΛ
with the linear in qΛ spin-oscillator quasi-potential energy U
X
qx φx (sΛ ) + U 0 (sΛ ),
U (sΛ ; qΛ ) = 2µη
(1.4)
x∈Λ
where ψ0Λ (qΛ ) =
Q
x∈Λ
1
ψ0 (qx ), ψ0 (q) = (µπ −1 ) 4 exp{− µ2 q 2 }, is the ground state of the free
oscillator Hamiltonian, that is the first term in the right-hand side of (1.1) with η = 0. We
prove the maximal ordering in the corresponding systems (provided some simple conditions
on U 0 and φx are satisfied): magnetizations MΛl , l = 1, 3, QΛ are non-zero. No other ground
states are known with such the property.
Gibbsian ground states were introduced by Kirkwood and Thomas in [5] in XZ spin- 12 models
with Hamiltonians (linear in S 1 ) that include the spin part of (1.1), i.e. the second and third
terms, with only Jx 6= 0 and the periodic boundary condition (this boundary condition is not
essential). Matsui in [6, 7] enlarged a class of spin- 12 XZ-type models in which Gibbsian states
exist. The method was further developed by Datta and Kennedy in [8]. An application of
classical spins systems for constructing of quantum states was given in [9].
In [10] we showed how to find VΛ for a given Gibbsian ground state and established existence
of lro’s in S 1 and S 3 for a wide class of the spin- 12 XZ models (see also [11, 12]). This reference
contains the most simple proofs of the existence of lro’s in ground states of quantum many-body
systems. A reader may find a review of the results concerning several quantum orders in it (see
also [13, 14, 15]).
The ground state in (1.3) can be represented in the following equivalent form
X 1
X
0
e− 2 U∗ (sΛ ) ψΛ
ΨΛ (qΛ ) =
(sΛ )ψ0Λ (qΛ + ηφΛ (sΛ )),
U∗ = U 0 − µη 2
φ2x .
(1.5)
sΛ
x∈Λ
orthonormality
of the basis (see the beginning of the third section) and the
From (1.5),
R
R
equalities ψ02 (q)dq = 1, ψ02 (q)qdq = 0 it follows that
hˆ
qx qˆy iΛ = η 2 hφx (SΛ3 )φy (SΛ3 )iΛ = η 2 hφx (σΛ )φy (σΛ )i∗Λ ,
hSx3 Sy3 iΛ = hσx σy i∗Λ ,
(1.6)
where σx (sΛ ) = sx and
−1
hφx (σΛ )φy (σΛ )i∗Λ = Z∗Λ
X
sΛ
e−U∗ (sΛ ) φx (sΛ )φy (sΛ ),
Z∗Λ =
X
e−U∗ (sΛ ) .
sΛ
Equalities in (1.6) reduce a calculation of averages in our quantum systems to a calculation of
averages indexed by a star in Ising models. For a short-range ferromagnetic potential energy U∗
the first Griffiths inequality holds: hσ[A] i∗Λ ≥ 0. As a result the following statement (principle)
is true.
Proposition 1. Let J0 ≥ 0 in (1.2) and ferromagnetic lro occur in the Ising model with the
ferromagnetic potential energy U∗ given by (1.5), that is hσx σy i∗Λ > 0 uniformly in Λ, then
ferromagnetic lro occurs in oscillator variables and S 3 in the quantum spin-oscillator system
with the Hamiltonian (1.1) and ground state ΨΛ in (1.3).
Usefulness of Gibbsian ground states is explained by comparative simplicity of a proof of
existence of lro. Our results show that Gibbsian ground states are expected to appear in many
quantum spin-oscillator systems with non-trivial interactions.
Our paper is organized as follows. In Section 2 we formulate our main results in a lemma
and theorem. In next sections we prove them.
4
2
T.C. Dorlas and W.I. Skrypnik
Main result
We establish that Gibbsian ground states exist for JA depending on qΛ if
VΛ = −
X
1
3
JA e− 2 WA (SΛ ) ,
WA (SΛ3 ) = U (SΛ3A ; qΛ ) − U (SΛ3 ; qΛ ),
(2.1)
A⊆Λ
3 , S3
where SΛ3A = (−SA
Λ\A ). VΛ determines an unbounded operator. Negative JA generate
positive functions VΛ and this enables us to prove the following lemma.
Lemma 1. Let VΛ be given by (2.1) and have a domain D(VΛ ), JA be bounded negative functions
and U in (2.1) coincide with U in (1.4) Then HΛ is positive definite and essentially self-adjoint
on the set (⊗ C2 )|Λ| ⊗S(R|Λ| )∩D(VΛ ) that contains (⊗ C2 )|Λ| ⊗C0∞ (R|Λ| ) and ΨΛ , given by (1.3),
is an eigenfunction of the Hamiltonian (1.1) with the zero eigenvalue and is its (unique) ground
state if JA ≤ 0 (if the uniform bound Jx ≤ J− < 0 holds).
Remark 1. If the functions JA are only negative then ΨΛ is the ground state of the self-adjoint
extension of the Hamiltonian preserving positive definiteness.
The most simple translation invariant short-range U 0 is ferromagnetic
U 0 (sΛ ) = −α
X
JA0 s[A] ,
JA0 ≥ 0,
(2.2)
A⊆Λ
where JA0 = 0 for odd |A|, α ≥ 0.
Theorem 1. Let all the conditions of Lemma 1 be satisfied, J0 ≥ 0, J0 (x; A) = 0 for even |A|,
0 ≥ J¯ and U 0 be given by (2.2). Then
J0 (0; 1), J0,1
I. For a sufficiently large β = (η 2 µ+α)J¯ > 1 there exist ground state lro’s in S 3 and oscillator
variables for d ≥ 2;
II. Let the positive constants C, Bj , j = 1, 2, independent of Λ, exist such that |φx (sΛ )| ≤ C
(j)
WA (sΛ ) =
X
j
|φjx′ (sA
Λ ) − φx′ (sΛ )| ≤ Bj ,
j = 1, 2,
x′ ∈Λ
where |A| = 2. Then ground state ferromagnetic lro in S 1 occurs in arbitrary dimension d.
If U 0 = 0 then VΛ from (2.1) is given by
VΛ = −
X
JA v[A] ,
vx = cosh ux + Sx3 sinh ux ,
ux = 2ηµφx (qΛ ).
A⊆Λ
This equality follows from the equalities
WA (SΛ3 ) = −2
X
ux Sx3 ,
3
eaS = cosh a + S 3 sinh a,
(S 3 )2 = I .
x∈A
In the simplest case the conditions of item II of Theorem 1 can be checked without difficulty.
Proposition 2. Let φx be linear in S 3 in (1.2) and ||J0 ||1 =
P
x
|J0 (x)| < ∞, where the sum-
mation is performed over Zd . Then the conditions of item II of Theorem 1 are satisfied.
Three Order Parameters in Quantum XZ Spin-Oscillator Models
5
Note that if one uses the Pauli matrices with 12 instead of the unity as matrix elements
then VΛ should be changed by adding to WA the number −|A| ln 2.
If one chooses the anti-ferromagnetic U 0 in (2.1), specifically,
X
X
U 0 (sΛ ) = µη 2
φ2x (sΛ ) + α
sx sy ,
α > 0,
x∈Λ
hx,yi∈Λ
where x, y are nearest neighbors, then it can be easily proved that the spin lro in the third
direction will be anti-ferromagnetic, generating a staggered magnetization (spins at the even
and odd sublattices take different values).
The interesting and important property of the Hamiltonians with VΛ given by (2.1) is that
they are simply related to generators of stationary Markovian processes (see, also, [16]). We
believe that it is possible to apply the same mathematical technique for proving existence of
order and phase transitions in equilibrium quantum systems and non-equilibrium stochastic
systems (see [17]).
3
Proof of Lemma 1
For our purpose it is convenient to pass to a new representation. It is determined by the Hilbert
space of sequences of functions F (qΛ ; sΛ ), sx = ±1, which are found in the expansion of the
vector F ∈ L2Λ mentioned at the beginning of the introduction, with the scalar product
XZ
F1 (qΛ ; sΛ )F2 (qΛ ; sΛ )dqΛ ,
(F1 , F2 ) =
sΛ
(F1 (qΛ ), F2 (qΛ ))0 =
X
F1 (qΛ ; sΛ )F2 (qΛ ; sΛ ),
sΛ
where the integration is performed over R|Λ| . Here we took into account the orthonormality of
the basis, i.e. the equality
Y
δsx ,s′x ,
(Ψ0Λ (sΛ ), Ψ0Λ (s′Λ ))0 = δ(sΛ ; s′Λ ) =
x∈Λ
where δs,s′ is the Kronecker delta. Let
hx = −∂x2 + µ2 (qx + ηφx (SΛ3 ))2 − µ,
h0x = −∂x2 + µ2 qx2 − µ = (−∂x + µqx )(∂x + µqx )
and
hΛ =
X
hx ,
h0Λ =
X
h0x .
x∈Λ
x∈Λ
(Tx (φ)F )(qΛ ) =
X
0
F (qx + ηφx (sΛ ), qΛ\x ; sΛ )ψΛ
(sΛ ),
sΛ
then
hx = Tx (φ)h0x Tx−1 (φ),
hΛ = TΛ h0Λ TΛ−1 .
Here we took into account that differentiation commutes with TΛ .
Our Hamiltonian is rewritten as follows
X
1
3
1
HΛ = hΛ +
JA P[A] ,
PA = S[A]
− e− 2 WA (SΛ ) .
A⊆Λ
6
T.C. Dorlas and W.I. Skrypnik
The remarkable fact is that the symmetric operator PA and the harmonic operator hx have both
common eigenvector ΨΛ with the zero eigenvalue (see Remark 3 in the end of the paper). Note
that the space of ground states of the operator hx (eigenfunctions with the zero eigenvalue) is
2|Λ| -fold degenerate since Sx3 is diagonal and the Laplacian is translation invariant. From (1.4)
and the definition of Tx (φ) if follows that Tx−1 (φ)ΨΛ is equal to ψ0 (qx ) multiplied by a function
independent of qx
X 1
0
(sΛ )ψ0Λ (qΛ\x + ηφΛ\x (sΛ\x ))ψ0 (qx ).
e− 2 U∗ (sΛ ) ψΛ
Tx−1 (φ)ΨΛ =
sΛ
Hence h0x Tx−1 (φ)ΨΛ = 0 and hx ΨΛ = Tx (φ)h0x Tx−1 (φ)ΨΛ = 0. The proof that ΨΛ is an eigenvector with the zero eigenvalue of PA is inspired by our previous paper [10]. For simplicity we will
omit qΛ in the expression for U in (1.3). Taking into consideration the equalities
0
0
(sΛ ) = sx ψΛ
(sΛ ),
Sx3 ψΛ
0
0 A
1
0
(sΛ\A , −sA ),
S[A]
ψΛ
(sΛ ) = ψΛ
(sΛ ) = ψΛ
we obtain
(ψ0Λ )−1 PA ΨΛ =
X
1
1
0
0
(ψΛ
(sΛ\A , −sA ) − e− 2 WA (sΛ ) ψΛ
(sΛ ))e− 2 U (sΛ )
sΛ
=
X
1
1
A
0
0
(sΛ\A , −sA )e− 2 U (sΛ ) − ψΛ
(ψΛ
(sΛ )e− 2 U (sΛ ) )
sΛ
=
X
1
A
1
A
0
(sΛ ) = 0.
(e− 2 U (sΛ ) − e− 2 U (sΛ ) )ψΛ
sΛ
Here we changed signs of the spin variables sA in the first term in the sum in sΛ .
Positive definiteness of the Hamiltonian follows from the following proposition.
Proposition 3. The operator −PA is positive definite on D(VΛ ).
Proof . VΛ is an operator of multiplication by infinite differentiable functions on each onedimensional spin subspace and its domain contains (⊗C2 )|Λ| ⊗ C0∞ (R|Λ| ).PThis domain coincides
|qx ||φx |.
with the direct sum of 2|Λ| copies of of L2 (R|Λ| , e|q|0 dqΛ ), where |q|0 =
x∈Λ
R
The scalar product in the Hilbert space L2Λ is given by (F1 , F2 ) = (F1 (qΛ ), F2 (qΛ )) dqΛ ,
where the integration is performed over R|Λ| . We have to show that −(PA F (qΛ ), F (qΛ ))0 ≥ 0.
Let us define the operator
1
1
3
3
PA+ = e 2 U (SΛ ) PA e− 2 U (SΛ ) .
0 that
It is not difficult to check on the basis ψΛ
1
3
1
PA+ = e− 2 WA (SΛ ) (S[A]
− I),
where I is the unit operator. Here we used the following equality
1
1
3A
3
1
1 − 2 U (SΛ )
= S[A]
e
.
e− 2 U (SΛ ) S[A]
For the operator PA+ we have
X
0
(PA+ F )(qΛ ; sΛ )ψΛ
(sΛ )
PA+ F =
sΛ
and
1
(PA+ F )(qΛ ; sΛ ) = −e− 2 WA (sΛ ) (F (qΛ ; sΛ ) − F (qΛ ; sA
Λ )).
Three Order Parameters in Quantum XZ Spin-Oscillator Models
7
It is convenient to introduce the new scalar product
XZ
3)
−U (SΛ
F1 (qΛ , sΛ )F2 (qΛ , sΛ )e−U (sΛ ) dqΛ
(F1 , F2 )U = (e
F1 , F 2 ) =
sΛ
=
Z
(F1 (qΛ ), F2 (qΛ ))0U dqΛ
=
Z
3
(e−U (SΛ ) F1 (qΛ ), F2 (qΛ ))0 dqΛ .
The operator PA+ is symmetric with respect to the new scalar product since
1
1
3
3
(PA+ F1 (qΛ ), F2 (qΛ ))0U = (PA e− 2 U (SΛ ) F1 (qΛ ), e− 2 U (SΛ ) F2 (qΛ ))0 .
(3.1)
It is not difficult to check that
X 1
A
e− 2 [U (sΛ )+U (sΛ )] (F (qΛ ; sΛ ) − F (qΛ ; sA
−(PA+ F (qΛ ), F (qΛ ))0U =
Λ ))F (qΛ ; sΛ )
sΛ
1 X − 1 [U (sΛ )+U (sA )]
Λ (F (q ; s ) − F (q ; sA ))2 ≥ 0.
=
e 2
Λ Λ
Λ Λ
2 s
(3.2)
Λ
Here we took into account that the function under the sign of exponent is invariant under
changing signs of the spin variables sA . From (3.1), (3.2) it follows that −PA ≥ 0.
P
JA PA is positive definite since for the negaFrom Proposition 3 it follows that HΛ1 =
A⊆Λ
tive JA we have
(HΛ1 F, F )
=
XZ
JA (qΛ )(PA F (qΛ ), F (qΛ ))0 dqΛ ≥ 0.
A⊆Λ
The fact that the Hamiltonian is essentially self-adjoint is derived from the following proposition (see example X.9.3 in [18]).
Proposition 4. The operator −∆ + µ2 x2 + V + (y, x), where (y, x) is the Euclidean scalar
n
P
∂
product of x = xj , j = 1, . . . , n with the constant vector y, ∆ =
∂xj , µ 6= 0 if y 6= 0, V is
j=1
2
the operator of multiplication by a non-negative function V (x) ∈ L2 (Rn , e−x dx), x2 = (x, x),
is essentially self-adjoint on S(Rn ) ∩ D(V ), i.e. the intersection of the Schwartz space and the
domain of V .
Proof . The following two inequalities are valid
µ2 ||x2 ψ|| ≤ ||(−∆ + V + µ2 x2 )ψ|| + 2µ2 n||ψ||,
(y, x) ≤ ǫx2 +
n 2
|y| ,
4ǫ
(3.3)
where µ is a real number, |y| = max |y|j , || · || is the scalar product in the Hilbert space of
j
square integrable functions. We tacitly assume that x2 means the operator of multiplication of
a squared variable. The first inequality follows from the inequalities (∂j = ∂x∂ j )
(−∆ + V + µ2 x2 )2 ≥ (µ2 x2 )2 − µ2 (∆x2 + x2 ∆),
−(∂j2 x2k + x2k ∂j2 ) = −(2xk ∂j2 xk + 2δj,k ) ≥ −2δj,k ,
n
X
−(∆x2 + x2 ∆) ≥ −
(∂j2 x2j + x2j ∂j2 ).
j=1
8
T.C. Dorlas and W.I. Skrypnik
Here we took into account that V is positive and [∂j , xk ] = ∂j xk − xk ∂j = δj,k . As a result
|y|2
2 2
−2
||(y, x)ψ|| ≤ a||(−∆ + V + µ x )ψ|| + b||ψ||,
a = µ ǫ,
b=n 2+
.
4ǫ
For µ 6= 0 number a can be arbitrary small and from the Kato–Rellich theorem [18] it follows
that the essential domain of −∆ + V + µ2 x2 + (y, x) coincides with the essential domain of
−∆+V +µ2 x2 . But the last one coincides with S(Rn )∩D(V ) [18, Theorem X.59]. Inequality (3.3)
is sufficient, also, for the proof of the proposition in the case µ = y = 0 (this is explained in
Example X.3 in [18]).
Since SΛ3 is a diagonal operator on (⊗C2 )|Λ| the operator
X
X
X
qx φx (SΛ3 )
qx2 + VΛ + 2ηµ2
∂x2 + µ2
−
x∈Λ
x∈Λ
x∈Λ
is the direct sum of 2|Λ| copies of the minus |Λ|-dimensional Laplacian plus the three functions
coinciding with the three functions in Proposition 4. From Proposition 4 it follows that this
2 |Λ| ⊗
operator is essentially self-adjoint on the set (⊗C2 )|Λ| ⊗ S(R|Λ| ) ∩ D(VP
Λ ) that contains (⊗C )
2
3
∞
|Λ|
2
φx (SΛ ) and the operator
C0 (R ). The same is true for the Hamiltonian since operator (ηµ)
x∈Λ
depending on SΛ1 in its expression are bounded.
The operators hΛ , HΛ1 are positive definite on the dense set (⊗C2 )|Λ| ⊗ S(R|Λ| ) ∩ D(VΛ ) which
is the essential set for HΛ . This implies that HΛ is positive definite on its domain D(HΛ ) and ΨΛ
is its ground state.
Proof of uniqueness. We have to establish that the symmetric semigroup PΛt , generated
by −HΛ , maps non-negative functions into (strictly) positive functions (increases positivity) and
this will imply that the ground state is unique (see Theorem XIII.44 in [18]). We will establish
this property with the help of a perturbation expansion. The kernel of the semigroup P1t ,
generated by hΛ + VΛ , is expressed in terms of the Feynman–Kac (FK) formula [18, 19]
Z
Zt
′
, s′Λ ) = δsΛ ,s′Λ PqtΛ ,q′ (dwΛ ) exp − VΛ+ (wΛ (τ ), sΛ )dτ ,
P1t (qΛ , sΛ ; qΛ
Λ
0
where VΛ+ (qΛ ; sΛ ) = VΛ (qΛ ; sΛ )+
P
[µ2 (qx +ηφx (sΛ ))2 −µ], PqtΛ ,q′ (dwΛ ) =
Λ
x∈Λ
Q
x∈Λ
Pqtx ,qx′ (dwx ) is the
conditional Wiener measure and wΛ (t) is the sequence of continuous paths. The semigroup P t
is represented as a perturbation series in powers of V0
X
1
V0 = −
JA S[A]
.
A⊆Λ
This series is convergent in the uniform operator norm [20] since V0 is a bounded operator. Its
perturbation expansion is given by
t
P =
X
n≥0
Pnt ,
Pnt
=
Z
0≤τ1 ≤τ2 ≤···≤τn ≤t
dτ1 · · · dτn P1τ1
n+1
Y
j=2
where τn+1 = t . We now use the following simple inequality
Z t
|VΛ+ (wΛ (τ ), sΛ )|dτ
0
τ −τj−1
(V0 P1 j
),
Three Order Parameters in Quantum XZ Spin-Oscillator Models
≤ V¯ (wΛ ) = |J|Λ
Z t"
(
¯0 + 2ηµ
exp αU
0
X
)
φ¯x |wx (τ )|
x∈Λ
+
X
9
#
(µ2 (|wx (τ )| + η φ¯x )2 − µ) dτ,
x∈Λ
¯0 = max U0 (sΛ ), |J|Λ = sup P |JA | , φ¯x = max |φx (sΛ )|. Let
where U
sΛ
sΛ
qΛ A⊆Λ
P¯1t (qΛ ; qΛ ) =
Z
¯
PqtΛ ,q′ (dwΛ )e−V (wΛ ) ,
V− = −J−
Λ
X
Sx1
x∈Λ
then it follows from the positivity of the the kernel P1t and V0 that
′
P t (qΛ , sΛ ; qΛ
, s′Λ ) ≥ P¯1t (qΛ ; qΛ )
X tn
n≥0
n!
V−n (sΛ ; s′Λ ).
(3.4)
Here we utilized the semigroup property of P¯ t , the inequalities ea ≥ e−|a| , V0 ≥ V− ,
τ −τj−1
(V0 P1 j
′
′
)V− (sΛ ; , s′Λ ).
, s′Λ ) ≥ P¯ τj −τj−1 (qΛ ; qΛ
)(qΛ , sΛ ; qΛ
Now, it is easily proved as in [10] that the matrix V− is irreducible. As a result there exists
a positive integer n such that and that V−n , has positive non-diagonal elements [21, 22]. Hence
the kernel in the left-hand side of (3.4) is positive.
4
Order parameters
In this section we will prove Theorem 1, i.e. occurrence of different types of lro in the considered
quantum systems. In the proof we will rely on the following basic theorem.
Theorem 2. Let the ferromagnetic short-range potential energy
P U of the classical Ising model
on the hyper-cubic lattice Zd with the partition function ZΛ = exp{−βU (sΛ )} be given by
sΛ
U (sΛ ) = −
X
(4.1)
ϕ(A)s[A] ,
A⊆Λ
where ϕ(A) ≥ 0, sx = ±1. Let, also, the uniform bound ϕ(x, y) ≥ ϕ¯ > 0 for nearest neighbors
x, y hold and ϕ(A) = 0 for A with odd number of sites. Then for a sufficiently large ϕβ
¯ >1
and the dimension d ≥ 2 there is the ferromagnetic lro, that is, for the Gibbsian two point spin
average the uniform in Λ bound holds
hσx σy iΛ > 0,
where σx (sΛ ) = sx and the magnetization (an order parameter) MΛ = |Λ|−1
P
sx is non-zero
x∈Λ
in the thermodynamic limit Λ → Zd .
The proof of this theorem is based on an application of the generalized Peierls principle
(argument). It will be given in the end of this section (see, also, [17, 26]. The next theorem is
the consequence of the basic theorem.
Proof of item I of Theorem 1. Condition (1.2) shows that
X
x∈Λ
φ2x (sΛ ) = |Λ| + 2
X
A⊆Λ
J2 (A)s[A] +
X
x∈Λ
X
x6∈A⊆Λ
2
J0 (x; A)s[A] .
10
T.C. Dorlas and W.I. Skrypnik
where J2 (x ∪ A) = J0 (x; A) and J2 (A) = 0 for odd |A|. The last term is equal to
X X
J02 (x; A) + 2
X
X
J0 (x; A1 )J0 (x; A2 )s[A1 ∆A2 ] ,
x∈Λ x6∈A1 ,A2 ⊆Λ
x∈Λ x6∈A⊆Λ
where s∅ = 1 and A1 ∆A2 = (A1 ∪
PA2 )\(A1 ∩ A2 ). Due to translation invariance of interaction
the first term is bounded by |Λ| J02 (0; A), where the summation is performed over Zd , and
A
this expression is finite since the interaction is short-range. Hence subtracting a finite constant
proportional to |Λ| from U∗ one sees that the result admits representation (4.1) with positive J∗
instead of ϕ such that J∗ (A) = 0 for odd |A| and for nearest neighbors x, y the inequality
¯ The basic theorem and Proposition 1 imply occurrence of ferromagnetic
J∗ (x, y) ≥ (2η 2 µ + α)J.
3
lro in S and oscillator lro.
0 (s )ψ (q ). Then (1.2) and the defiProof of item II of Theorem 1. Let ψΛ (qΛ ; sΛ ) = ψΛ
Λ 0Λ Λ
1
nition of S imply that
X
Sx1 Sy1 ΨΛ (qΛ ) =
1
e− 2 U (sΛ ;qΛ ) ψΛ (qΛ ; sx,y
Λ )=
sΛ
X
x,y
1
e− 2 U (sΛ
;qΛ )
ψΛ (qΛ ; sΛ ).
sΛ
Taking into account also the orthonormality of the basis one obtains
XZ
x,y
1
2
hSx1 Sy1 iΛ = ZΛ−1
e− 2 [U (sΛ ;qΛ )+U (sΛ ;qΛ )] ψ0Λ
(qΛ )dqΛ
sΛ
=
ZΛ−1
X
e
η2 µ
4
P
x′ ∈Λ
2
(φx′ (sΛ )+φx′ (sx,y
Λ ))
α
x,y
e− 2 [U0 (sΛ
)+U0 (sΛ )]
sΛ
−α
B
2 0
≥e
ZΛ−1
X
e
η2 µ
4
P
x′ ∈Λ
2
(φx′ (sΛ )+φx′ (sx,y
Λ ))
e−αU0 (sΛ ) .
sΛ
We also have
X
2
(φx′ (sΛ ) + φx′ (sx,y
Λ ))
x′ ∈Λ
=
X
[3φ2x′ (sΛ ) + φ2x′ (sx,y
Λ )] + 2
x′ ∈Λ
x′ ∈Λ
=
X
4φ2x′ (sΛ ) +
x′ ∈Λ
≥4
φx′ (sΛ )[−φx′ (sΛ ) + φx′ (sx,y
Λ )]
X
X
[−φ2x′ (sΛ ) + φ2x′ (sx,y
Λ )] + 2
x′ ∈Λ
X
X
φx′ (sΛ )[−φx′ (sΛ ) + φx′ (sx,y
Λ )]
x′ ∈Λ
φ2x′ (sΛ ) − B2 − 2CB1 .
x′ ∈Λ
This yields
hSx1 Sy1 iΛ
2
B
− η 4µ (B2 +2B1 ) − α
2 0
≥e
e
ZΛ−1
X
η2 µ
e
P
x′ ∈Λ
φ2x′ (sΛ )
e−αU0 (sΛ )
sΛ
2
− η 4µ (B2 +2CB1 )
=e
−α
B
2 0
e
.
Proof of Proposition 2. It is obvious that |φx (sΛ )| ≤ ||J0 ||1 and that
′
′
′
′
|φx′ (sx,y
Λ ) − φx′ (sΛ )| = | − 2[sy J0 (y − x ) + sx J0 (x − x )]| ≤ 2[|J0 (y − x )| + |J0 (x − x )|].
Three Order Parameters in Quantum XZ Spin-Oscillator Models
11
(1)
As a result Wx,y (sΛ ) ≤ 4||J0 ||1 . Further
"
#2
X
x,y
2
2
′
′
′
|φx′ (sΛ ) − φx′ (sΛ )| =
J0 (z − x )sz − sy J0 (y − x ) − sx J0 (x − x )
z∈Λ\(x,y)
"
#2
X
−
J0 (z − x′ )sz + sy J0 (y − x′ ) + sx J0 (x − x′ )
z∈Λ\(x,y)
X
= −4
J0 (z − x′ )sz (sy J0 (y − x′ ) + sx J0 (x − x′ ))
z∈Λ\(x,y)
X
|J0 (z − x′ )|(|J0 (y − x′ )| + |J0 (x − x′ )|) ≤ 4||J0 ||1 (|J0 (y − x′ )| + |J0 (x − x′ )|).
≤4
z∈Λ
(2)
Hence Wx,y (sΛ ) ≤ 8||J0 ||21 .
1
Proof of Theorem 2. Let χ±
x = 2 (1 ± σx ) then one obtains
−
4hχ+
x χy iΛ = 1 + hσx iΛ − hσy iΛ − hσx σy iΛ .
Since the systems are invariant under the transformation of changing signs of spins the third
and the second terms in the right-hand side of last equality are equal to zero and
−
hσx σy iΛ = 1 − 4hχ+
x χy iΛ .
Hence if
−
hχ+
x χy iΛ <
1
4
(4.2)
then the ferromagnetic lro occurs, i.e.
hσx σy iΛ ≥ a > 0,
(4.3)
where a is independent of Λ. If one succeeds in proving that there exists a positive function E0 (β)
and positive constants a, a′ independent of Λ such that
′ E0 (β)
−
hχ+
x χy iΛ ≤ a e
(4.4)
and proves that E0 is increasing at infinity then (4.2), (4.3) will hold for a sufficiently large
inverse temperature β. The Peierls principle reduces the derivation of (4.4) to the derivation of
the contour bound.
Peierls principle. Let the contour bound hold
h
Y
−
−|Γ|E
,
χ+
x χy iΛ ≤ e
(4.5)
hx,yi∈Γ
where h·iΛ denotes the Gibbs average for the spin system confined to a compact domain Λ,
Γ is a set of the nearest neighbors, adjacent to the (connected) contour, i.e. a boundary of the
connected set of unit hypercubes centered at lattice sites. Then (4.4) is valid with E0 = a′′ E,
where a′′ is a positive constant independent of Λ.
12
T.C. Dorlas and W.I. Skrypnik
Proof of contour bound. Bricmont and Fontain derived the contour bound for the spin systems with the potential energy (4.1) with the help of the second Griffiths [23] and Jensen
inequalities [24] (see also [25, 26])
Z
Z
f
hσ[A] σ[B] iΛ[Γ] − hσ[A] iΛ[Γ] hσ[B] iΛ[Γ] ≥ 0,
e dµ ≥ exp
f dµ ,
where dµ is a probability measure on a measurable space. Their proof starts form the inequality
β
β
β
β
− 2 σx σ y 2 σx σ y + −
−
− 2 σ x σy
−
χ+
e
χx χy ≤ e− 2 σx σy χ+
.
x χy = e
x χy ≤ e
As a result (β ′ = ϕβ)
¯
′
h
Y
−
χ+
x χy iΛ
− β2
≤ he
P
σx σy
iΛ = he
hx,yi∈Γ
β′
2
P
hx,yi∈Γ
σ x σy
i−1
Λ[Γ]
hx,yi∈Γ
′
− β2
≤e
P
hσx σy iΛ[Γ]
hx,yi∈Γ
= e−EΓ ,
where h·, ·iΛ[Γ] is the average corresponding to the potential energy
UΓ (qΛ ) = U (sΛ ) +
ϕ¯ X
sx sy .
2
hx,yi∈Γ
In the last line we applied the Jensen inequality. From the second Griffiths inequality it follows
that the average hσx σy iΛ[Γ] is a monotone increasing function in ϕA . So, in the potential energy
determining this average we can put ϕA = 0, except A = hx, yi and leave the coefficient ϕ¯ in
front of the bilinear nearest-neighbor pair potential in (4.1) without increasing the average. This
leads to
′ X
X
β′
β′
−1 β
′
Z2 (β) =
e 2 s1 s2 .
hσx σy iΛ[Γ] ≥ hσσ i = Z2
s1 s2 e 2 s1 s2 ,
2
s1 ,s2 =±1
s1 ,s2 =±1
That is,
E = 2−1 β ′ hσσ ′ i
EΓ ≥ |Γ|E,
or
E = β(e2
−1 β ′
− e−2
−1 β ′
)(e2
−1 β ′
+ e−2
−1 β ′
Here we used in the denominator the inequality e−2
if β ′ tends to infinity. This implies (4.5).
5
′
)−1 ≥ 2−1 β ′ (1 − e−β ).
−1 β ′
≤ e2
−1 β ′
. Obviously, E tends to infinity
Discussion
We showed that in the considered lattice spin-boson models with JA ≤ 0 ground states are
Gibbsian and the ground state averages for special observables are reduced to averages in classical
Ising models. This means that existence of ground states order parameters is connected with
existence of order parameters in the associated Ising models and that a breakdown of symmetries
in the quantum systems is determined by a breakdown of symmetries in Ising models. We
considered the free boundary conditions implying that for the cases of the perturbation VΛ ,
Three Order Parameters in Quantum XZ Spin-Oscillator Models
13
considered in the two theorems, the ground state averages of qx , Sx3 are zero, that is hˆ
qx iΛ = 0,
hSx3 iΛ = 0 if the associated Ising potential energy is an even function. In order to make such the
averages non-zero (explicit symmetry breaking) one has to introduce special boundary conditions
(quasi-averages) which have to single out pure Gibbsian states in the associated Ising models. It
is known [28] that for the two-dimensional ferromagnetic Ising nearest-neighbor model there are
two boundary conditions which generate pure states and that every other state is a convex linear
combination of these two states. A discussion of a construction of ground states in lattice spin
and fermion quantum systems with an explicit symmetry breaking a reader may find in [29].
Remark 2. Translation invariance means that
Jx1 ,...,xn = J0,x2 −x1 ,...,xn −x1 ,
J0 (x; x1 , . . . , xn ) = J0 (0; x1 − x, . . . , xn − x)
where J, J0 are symmetric functions. The short-range character of interaction means that
X
X
|J0 (x; A)| < ∞.
|Jx,A | < ∞,
max
max
x
A
x
A
Remark 3. If only one-point sets are left in the sum for VΛ then the expression for HΛ can be
rewritten in the following way
X
HΛ =
Hx .
x∈Λ
The property of the ground state ΨΛ to be a ground state with the zero eigenvalue of a local
Hamiltonian Hx was found earlier for special isotropic anti-ferromagnetic Heisenberg chains with
valence bond ground state in [27].
References
[1] Kobayashi K., Dynamical theory of the phase transition in KH2 PO4 -type ferroelectric crystals, J. Phys.
Soc. Japan 24 (1968), 497–508.
[2] Villain J., Stamenkovic S., Atomic motion in hydrogen-bond ferroelectrics, Phys. Stat. Sol. 15 (1966), 585–
596.
[3] Kurbatov A., Plechko V., Exactly solvable model with order-disorder phase transition for ferroelectrics,
Teoret. Mat. Fiz. 26 (1976), 109–116.
[4] Bogolyubov N.N. Jr., Brankov J.G., Kurbatov A.M., Tonchev N.C., Zagrebnov V.A., Method of approximating Hamiltonian in statistical physics, Sophia, 1981.
[5] Kirkwood J., Thomas L., Expansions and phase transitions for the ground state of quantum Ising lattice
systems, Comm. Math. Phys. 88 (1983), 569–580.
[6] Matsui T., A link between quantum and classical Potts models, J. Statist. Phys. 59 (1990), 781–798.
[7] Matsui T., Uniqueness of the translationally invariant ground state in quantum spin systems, Comm. Math.
Phys. 126 (1990), 453–467.
[8] Datta N., Kennedy T., Expansions of one quasiparticle states in spin-1/2 systems, J. Statist. Phys. 108
(2002), 373–399, cond-mat/0104199.
[9] Goderis D., Maes C., Constructing quantum dissipations and their reversible states from classical interacting
spin systems, Ann. Inst. H. Poincar´e Phys. Th´eor. 55 (1991), 805–828.
[10] Dorlas T., Skrypnik W., On quantum XZ 1/2 spin models with Gibbsian ground states, J. Phys. A: Math.
Gen. 37 (2004), 66623–6632.
[11] Skrypnik W., Order parameters in XXZ-type spin
2 (2006), 011, 6 pages, math-ph/0601060.
1
2
quantum models with Gibbsian ground states, SIGMA
[12] Skrypnik W., Quantum spin XXZ-type models with Gibbsian ground states and double long-range order,
Phys. Lett. A 371 (2007), 363–373.
14
T.C. Dorlas and W.I. Skrypnik
[13] Messager A., Nachtergaele B., A model with simultaneous first and second order phase transitions, J. Statist.
Phys. 122 (2006), 1–14, cond-mat/0501229.
[14] van Enter A.C.D., Shlosman S.B., First-order transitions for n-vector models in two and more dimensions;
rigorous proof, Phys. Rev. Lett. 89 (2002), 285702, 3 pages, cond-mat/0205455.
[15] van Enter A.C.D., Shlosman S.B., Provable first-order transitions for liquid crystal and lattice gauge models
with continuous symmetries, Comm. Math. Phys. 255 (2005), 21–32, cond-mat/0306362.
[16] Thomas L.E., Quantum Heisenberg ferromagnets and stochastic exclusion processes, J. Math. Phys. 21
(1980), 1921–1924.
[17] Skrypnik W., Long-range order in nonequilibrium systems of interacting Brownian linear oscillators,
J. Statist. Phys. 111 (2003), 291–321.
[18] Reed M., Simon B., Methods of modern mathematical physics, Vols. II, IV, Academic Press, 1975.
[19] Ginibre J., Reduced density matrices of quantum gases. I. Limit of infinite volume, J. Math. Phys. 6 (1964),
238–251.
[20] Kato T., Perturbation theory for linear operators, Springer-Verlag, 1966.
[21] Gantmacher F.R., Matrix theory, 4th ed., Nauka, Moscow, 1988.
[22] Gantmacher F.R., Applications of the theory of matrices, Interscience Publ., New York, 1959.
[23] Kelly D., Sherman S., Inequalities on correlations in Ising ferromagnets, J. Math. Phys. 9 (1968), 466–483.
[24] Bricmont J., Fontaine J.-R., Correlation inequalities and contour estimates, J. Statist. Phys. 26 (1981),
745–753.
[25] Fr¨
ohlich J., Lieb E., Phase transitions in anisotropic lattice spin systems, Comm. Math. Phys. 60 (1978),
233–267.
[26] Skrypnik W., Long-range order in linear ferromagnetic oscillator systems. Strong pair quadratic n-n potential, Ukrainian Math. J. 56 (2004), 964–972.
[27] Affleck I., Kennedy T., Lieb E., Tasaki H., Valence bond ground states in isotropic quantum antiferromagnets, Comm. Math. Phys. 115 (1988), 477–528.
[28] Izenman M., Translation invariance and instability of phase coexistence in the teo dimensional Ising system,
Comm. Math. Phys. 73 (1980), 84–92.
[29] Koma T., Tasaki H., Symmetry breaking and finite-size effects in quantum many-body systems, J. Statist.
Phys. 76 (1994), 745–803, cond-mat/9708132.